[
https://issues.apache.org/jira/browse/PARQUET-220?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Konstantin Shaposhnikov reopened PARQUET-220:
---------------------------------------------
Sorry for late response, I was away from my computer on holidays.
I've created a new pull request that hopefully addresses the issue with Hadoop
1.x: https://github.com/apache/incubator-parquet-mr/pull/163
The fix moves the check and logging to ContextUtil getCounter() method. Context
class doesn't need to be TaskInputOutputContext, it just needs to have
getCounter(String, String) method.
I've included your changes to the pull request (pom.xml and new deprecated
methods) as well.
> Unnecessary warning in ParquetRecordReader.initialize
> -----------------------------------------------------
>
> Key: PARQUET-220
> URL: https://issues.apache.org/jira/browse/PARQUET-220
> Project: Parquet
> Issue Type: Improvement
> Affects Versions: parquet-mr_1.6.0
> Reporter: Konstantin Shaposhnikov
> Fix For: 1.6.0
>
>
> When reading a parquet file using spark 1.3.0 lots of warnings are printed in
> the log:
> {noformat}
> WARNING: parquet.hadoop.ParquetRecordReader: Can not initialize counter due
> to context is not a instance of TaskInputOutputContext, but is
> org.apache.hadoop.mapreduce.task.TaskAttemptContextImpl
> {noformat}
> I have checked the source of ParquetRecordReader and found that while it
> checks for context to be TaskInputOutputContext it seems to never actually
> rely on this fact.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
